Long-term Reference Picture Signaling via Predictive POC Coding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for signaling long-term reference pictures in video coding are either not error-resilient or inefficient, and changing a Clean Random Access (CRA) picture to a Broken Link Access (BLA) picture requires unnecessary bit changes and waste.
Innovation Solution
The proposed solution involves predictive coding of most significant bits (MSBs) of picture order count (POC) values for long-term reference pictures and optimizing the signaling of random access point pictures, including Instantaneous Decoding Refresh (IDR), Clean Random Access (CRA), and Broken Link Access (BLA) pictures, to improve coding efficiency and error resilience.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If full POC values are coded for long-term reference pictures, then accuracy is improved, but bit rate increases and coding efficiency decreases
Solution Approach 1:
The POC value is segmented into two parts: least significant bits (LSBs) and most significant bits (MSBs). The LSBs are coded explicitly while the MSBs are predicted from a reference POC value. This segmentation allows the system to maintain sufficient precision for identifying reference pictures while significantly reducing the number of bits that need to be transmitted, thereby improving coding efficiency.
Solution Approach 2:
A predicted POC value is introduced as an intermediary element. The decoder uses a reference POC value (from the current or previous picture) to predict the MSBs of the long-term reference picture's POC value. This predicted value serves as a mediator that enables accurate reconstruction of full POC values without transmitting them directly, thus improving compression efficiency.
2Ease of operation
If CRA pictures are used for random access, then decoding simplicity is improved, but error resilience deteriorates
Solution Approach 1:
The system dynamically adapts the picture type between CRA and BLA based on the presence of long-term reference pictures and the current picture's reference status. When LTROPs are present and the current picture is referenced, the system switches to BLA mode, which provides enhanced error resilience by allowing proper reference picture management across random access points, while maintaining CRA simplicity when these conditions are not met.
3Reliability
If BLA pictures are used for random access, then error resilience is improved, but bitstream complexity increases
Solution Approach 1:
The system changes key parameters in the picture header and reference picture list management when switching between CRA and BLA modes. Specifically, it modifies the `no_output_to_prior_pics_flag`, updates reference picture list structures, and adjusts POC value handling. These parameter changes enable the system to achieve enhanced error resilience while controlling bitstream complexity through targeted modifications rather than complete redesign.
4Quantity of substance
If predictive coding of MSBs is used, then bit rate is reduced, but decoding complexity increases
Solution Approach 1:
The decoder performs preliminary actions by first identifying the reference picture (current or previous) and retrieving its POC value before decoding the difference. This preliminary retrieval and preparation of the reference POC value simplifies the subsequent decoding process, as the decoder only needs to add the difference to the pre-fetched reference value rather than searching or complex computation during the actual decoding phase.
Data Source
AI summary
In one example, a device for decoding video data includes a video decoder configured to decode a value representative of a difference between most significant bits (MSBs) of a reference picture order count (POC) value and MSBs of a long-term reference picture (LTRP) POC value, wherein the reference POC value corresponds to a picture for which data must have been received in order to properly decode a current picture, determine the MSBs of the LTRP POC value based on the decoded value and the reference POC value, and decode at least a portion of the current picture relative to the LTRP based at least in part on the LTRP POC value. The picture for which data must have been received in order to properly decode a current picture may correspond to the current picture itself or a most recent random access point (RAP) picture.